Chris Lamoureux's Research Page

• Published Papers



• Unpublished working papers:
  • The Gibbs Posterior and Parametric Portfolio Choice New paper, posted March 6, 2026. I wrote this paper to address several things that bothered me after completing the 2024 RAPS paper with Huacheng. Also, I consider myself a Bayesian and I like to integrate estimation and decsion-making. The loss function that underpins the decision should be the same that does the estimation. I came across Bissiri, Holmes, and Walker's 2016 "A general framework for updating belief distributions," (JRSS-B) and light bulbs started flashing. This contained the answer to a question I've been asking for 10 years: What is a Bayesian to do when he has a policy function and data but doesn't want to write down a likelihood? I knew that empirical likelihood methods could handle this problem, but I was skeptical of their generality. So this is how this paper addresses concerns that lingered after the RAPS paper:
    1. I know my investor's utility function. Why am I choosing the specification using maxmin? Of course this is the multi-prior solution to ambiguity, but it seems ad-hoc in this application. I would like to integrate over estimation risk in my decision making.
    2. The need for out-of-sample regularization seems wasteful. I would prefer a way to regularize in-sample.
    3. The bootstrap makes assumptions about the data generating process that are inconsistent with the agnostic approach to the data of Brandt, Santa-Clara, and Valkanov's algorithm. It also violates the likelihood principle. I want my decision to rely only on observed data.
    4. Finally, as a Bayesian I would like a procedure that updates my prior belief given the data in a coherent way, rationally linking my inference to my decision making.
    I hope that this paper directly addresses and solves all of these concerns on Huacheng and my earlier paper. I especially like the mechanism by which the Gibbs sampler produces posteriors on non-linear functions of model / policy hyperparameters. Indeed, this is why I became a Bayesian. We have a natural way to analyze the consequences of parameter uncertainty on things we care about.
    
    
  • ``An Empirical Assessment of Characteristics and Optimal Portfolios,'' (with Huacheng Zhang). New revision posted February 1, 2024. This pdf file includes the Internet Appendix. This paper is forthcoming in the Review of Asset Pricing Studies. I am a huge fan of Brandt, Santa-Clara and Valkanov's 2009 paper, "Parametric Portfolio Policies: Exploiting characteristics in the Cross-Section of Equity Returns," ( Review of Financial Studies). The original motivation for this this paper was to ascertain whether the month-of-the-year effect, which was originally described by Jegadeesh, and more recently highlighted by Heston and Sadka, and other characteristics were robust enough to a) remain useful to an investor who cared about higher moments than characterized by the Sharpe ratio; and/or b) were useful individually and jointly. Once Huacheng and I started analyzing things another, perhaps more important question arose--that of overfitting and estimation risk. For example, a power utility investor with a coefficient of relative risk aversion, g, of 2 would pay to avoid the optimal portfolio, in light of its very poor out-of-sample performance. Much like empirical mean-variance optimal portfolios estimation risk was more important than exploiting in-sample patterns. I thought a natural way to regularize the estimation would be to ramp up the coefficient of risk aversion used to estimate portfolio weights. This worked quite well. Our g = 2 power utility investor should select portfolios optimal for a power utility investor with a coefficient of relative risk aversion, g = 3. This produced a portfolio that has excellent out-of-sample properties. In a sense we discovered something that is either trivial or subtle, that estimation risk lives in the portfolio variance. Thus any strategy that reduces portfolio variance is apt to lower estimation risk. We had to use a 3-stage sample design because we learned about overfitting by looking at the portfolio's performance in a second (i.e., out-of-sample) sample period. We selected the optimal regularization (being the increment to the investor's g to use in optimization), characteristic set, and protocol (being updating versus rolling). But this is itself optimally chosen from the "out-of-sample period." So true out-of-sample evaluation requires a third sample period. We selected the optimal across bootstrap synthetic samples using max-min. Which specification had the highest minimum certainty equivalent return across the bootstapped samples. We also found that the characteristics were synergistic in the sense that beating the market required at least three and the gains from adding characteristics were more than additive. Momentum, unconditionally is attractive but down-weighting momentum large-stocks and over-weighting small momentum stocks, even more so. Also to my surprise idiosyncratic volatility was the most robust and dominant characteristic. Underweighting high-vol stocks resulted in higher mean returns and lower variances. However, we also find evidence that the gains from characteristic weight tilts vanish after the year 2000.
  
  
• ``Dimensions of Limits to Arbitrage: Evidence from Coupon Spreads and Repo Specials in the 10-Year US Treasury Market'' (with George Theocharides). New revision posted July 29, 2013. First posted November 30, 2012.
  • Slides.
  
  
• ``Public Information and Stale Limit Orders: The Evidence'' (with Qin Wang). New draft posted August 11, 2015. First posted October 2, 2012.
  
  
• ``There and Back Again: A Stock's Tale.'' New revision posted March 27, 2017. (First posted June 28, 2012.)
  
  
  • I have never been a huge fan of case studies, since there is a lack of statistical power--the analysis cannot integrate over noise and possibly other confounding factors. However, in teaching and developing intuition I find that I am drawn to case studies because they allow a focused theoretical analysis. An example of this is the storied 3Com-Palm spin-off, which focuses attention on the cost of borrowing shares for the purpose of establishing and maintaining a short position. Another example is the strange case of Entremed, that Huberman and Regev (2001 JF) analyze. In this case the stock reacted to a New York Times story that contained old information (it had been published in Nature six months beforehand). My sense of this case is that the original Nature article may have been like the proverbial tree falling in the forest.
    
    My case study looks at the delisting of Maxim Integrated Products in 2007. Maxim was a $9 billion S&P 500 company, delisted for options backdating, and failure to restate its financials. This was a punitive regulatory delisting unrelated to the company's business. The stock trading moves from Nasdaq NMS to the lowest tier of the Pink Sheets (OTC PINK), its options are delisted, it is removed from the S&P 500 Index, the Nasdaq 100, the Russell 1000, and the Philadelphia Semiconductor Index. Despite all of this, its trading dynamics on the Pink Sheets and its ownership structure are virtually unchanged. I think this is interesting because it shows that the institution of where (and how) a company's shares are traded is not important per se. Whether a stock has listed options is not important per se as concerns the quality of the market. By quality of market, I mean the speed with which information is incorporated into the stock's price as well as price pressure and the bid-ask spread.
    
    The delisting was largely a non-event because market participants, especially institutions who use risk capital to make a market in the stock, expected that the company would soon relist. Three and a half months after delisting Maxim's management announced that their relisting would be delayed. Although this event was not associated with any institutional changes (i.e., the shares continue to trade on the Pink Sheets for another nine months), trading dynamics are adversely affected. Information assimilation slows down, trading volume, analyst following, and media attention all drop. Furthermore there is now significantly higher price pressure. Almost a year after delisting the company comes into compliance with SEC regulations and its shares and options are re-listed. Trading metrics return to pre-delisting levels.
    
    I think this separation of ``bad news'' from a regulatorily-induced delisting shows that investor attention is more important than institutions in affecting market quality. We need a case study to disentangle these two because they are naturally closely integrated. For example when options are listed on a company it is because there is a lot of trading and volatility in the stock. The listing of the options is not exogenous. When a company's shares are delisted it is almost always an absorbing barrier--the end game for those shares. But this paper shows that it's not the delisting that matters, rather it is the belief that this is a permanent state.
  
  
  
• ``Measuring Private Information in a Specialist Market,'' (with Qin Wang). New revision posted October 13, 2014. First posted July 3, 2012.
  
  
• ``Forecasting the Yield Curve Prior to the Global Financial Crisis: An Assessment of the Cox, Ingersoll, and Ross Model,'' (with Ken Roskelley). New revision posted June 8, 2017. (First posted August 8, 2005. This paper's earlier title was: ``Estimating and Testing Arbitrage Models without Adding an Error Model: An Application to Cox, Ingersoll, and Ross.'' )
  
  
• ``Costs of Capital and Public Issuance Choice,'' (with Ali Nejadmalayeri). New revision posted September 13, 2011. (This paper replaces an earlier paper entitled, ``Capital Market Conditions and Public Issuance Choice.'')
  
  
• ``Information in Option Prices and the Underlying Asset Dynamics,'' (with Alex Paseka). New revision posted October 30, 2009. (First posted on June 24, 2004.)
  
  
• ``Microstructure with Multiple Assets: An Experimental Investigation into Direct and Indirect Dealer Competition'' with Chuck Schnitzlein. (updated March 18, 2003).
  
  
• ``To Err is Human - But Dealers' Abilities to find non-Dominated Strategies Depends on Market Transparency'' with Chuck Schnitzlein. (updated June 7, 2001).
  
  
• ``Variations in Stock Returns: Asymmetries and Other Patterns'' with Sunil Panikkath. (May 1994).

Discussions and Other materials.

• Slides for my discussion of ``Tax-Timing Options and the Demand for Idiosyncratic Volatility,'' by Oliver Boguth and Luke C.D. Stein. (Tucson, April 29, 2017).
• Slides for my discussion of ``Flow and Stock Effects of Large-Scale Treasury Purchases,'' by Stefania D'Amico and Thomas King. (St. Louis, June 30, 2011).

Data, Programs and Other Resources.

• The original data used in my JF paper with Doug Witte (on Bayesian estimation of affine models).
• F0rtran code used to draw parameters and state variables for that paper.